The Short Version
- AI can help people name feelings, organize thoughts, prepare for hard conversations, and see another angle.
- The same tool can also sound caring, confident, and authoritative when it is really predicting helpful-sounding language from patterns.
- If you use AI for therapy or mental health support, build skepticism into the tool before you rely on it.
- Use the custom instructions below, test them in a fresh chat, and only save them if the answers are more useful.
AI for therapy is already common
About 1 in 8 people ages 12 to 21 are using AI chatbots for mental health advice. For adults ages 18 to 29, it is 28%. And in that 12-to-21 group, more than 93% of users said the advice was helpful.
I am not surprised people find it helpful. AI is available all the time. It answers instantly. It is basically free. It does not get tired of your questions. That is useful. But useful does not automatically mean safe, private, or accurate.
Copy/Paste Custom Instructions
When I ask for advice: Do not simply validate me. Separate facts, assumptions, feelings, and interpretations. Give more than one plausible interpretation. Tell me what you might be missing. Do not imitate intimacy. Help me leave the conversation with more agency.
Paste these into ChatGPT Custom Instructions, Claude preferences or project instructions, Gemini saved instructions if available, or the start of a new chat. If the tool ignores the instructions, start a fresh chat and paste them again.
Why these instructions matter
AI is built to mimic human communication. It can sound caring. It can sound confident. It can sound authoritative.
Ask it about medicine, and it can play a realistic doctor. Ask it about depression, marriage, addiction, grief, loneliness, or parenting, and it can play the part of someone wise and caring.
That can be useful. It can also be misleading if you forget what the tool is. Underneath, AI is predicting helpful-sounding language based on patterns. It can be a thinking partner. It should not become your answer machine.
The goal is to make the answer more useful. Less blind validation. More critical thinking. More separation between facts, feelings, assumptions, and interpretations.
Think carefully about privacy
Your private life may be stored, reviewed, searched, connected to other things you have said, leaked, subpoenaed, or used later in ways you never imagined.
This does not mean nobody should ever use AI for therapy or mental health support. It means you should understand the tradeoff before you pour your mental health, family conflict, addiction, trauma, marriage, or child's struggles into a system that does not belong to you.
Test Before You Save It
Do not take my word for it. Test the difference before saving these instructions into your AI rules.
- Open a new chat without the custom instructions.
- Ask the test prompt below and read the answer.
- Open a second new chat.
- Paste the custom instructions first, then ask the same prompt again.
- If the second answer helps you think more clearly, consider saving the instructions.
For parents
Your kid may already be using AI this way. My ideal is simple: kids should not use AI alone. They should be one hundred percent supervised, no exceptions.
But the numbers tell us some kids are already using AI for mental health advice. If you find out your kid is one of them, do not make secrecy the lesson. Start with curiosity.
What are you asking it? What is it telling you? Does it ever challenge you? Does it make you feel more capable or more dependent?
Bring it into the open. Help your kid understand what these systems are and are not. The skill is not blind trust, and it is not blind fear. The skill is skepticism.
Sources
- Brown University School of Public Health: One in eight adolescents and young adults use AI chatbots for mental health advice
- KFF: Tracking Poll on Health Information and Trust: Use of AI For Health Information and Advice
- Pew Research Center: Teens, Social Media and AI Chatbots 2025
- Stanford HAI: Exploring the Dangers of AI in Mental Health Care
- Brown University: AI chatbots systematically violate mental health ethics standards
- Google Gemini Apps Privacy Hub: Gemini Apps privacy information